Coalition Trading in Spain: Explaining State-wide Parties' Government Formation Strategies at the Regional Level

In compounded polities, political parties need to co-ordinate their governing strategies across the central and the regional tiers of government. The coalition game is also more complex in these contexts, especially for the state-wide parties (SWP) that alternate in central office, as their need for cohesiveness across party levels is higher and the quest for vertical congruence across institutional levels is more pressing. This article aims at identifying the influence of the SWP situation at the central level on their government formation strategies at the regional level. By examining the Spanish case, we show that multi-level dynamics provide parties with incentives to trade government formation strategies at different levels. Particularly, when the SWP in the central government has a minority status, parliamentary support is more likely to be given to the investiture of non-state-wide parties (NSWP) in those regions where the latter are stronger. Furthermore, we observe that, as political decentralization advanced, SWP seem to have been more inclined to adopt co-operative behaviour towards NSWP. Overall, our findings suggest that SWP's strategies are not simply interconnected across levels but rather regional government formation decisions are hierarchically subject to central-level considerations.


Introduction
The bulk of the literature on coalitions has traditionally focused on national governments. So far, very few studies have shifted their attention to the regional level (notable exceptions include Colomer and Martínez, 1995;Downs, 1998;Bäck, 2003;Reniu, 2005;Ş tefuriuc, 2007, 2009a, 2009bHough and Verge, 2009;Rodríguez Teruel et al., 2010). Interestingly, the regional level provides researchers with the opportunity to examine coalition formation within a single country while dealing prefer to govern alone, stay in opposition, or reach some kind of agreement with NSWP in the form of a coalition cabinet or an opposition -government collaboration.
The remainder of the article is structured as follows. The second section briefly introduces Spanish regional governments and SWP's participation in them. The third section specifies our hypotheses while the fourth describes the data and methods which drive the empirical analysis. The fifth section discusses the main findings in the light of our hypotheses. The final section highlights the contribution of the article and sets an agenda for further research.

State-wide Parties and Regional Governments in Spain
As Detterbeck and Hepburn (2010: 124) remind us, "parties are struggling as much with the implications of state structural change resulting from multi-level governance in their internal organization, as they are externally in their responses to new divisions of state powers". State decentralization processes impact on the most relevant dimensions of party organization, namely distributional conflicts over resources, the definition of platforms, and electoral and governing strategies (Downs, 1998;Chhibber and Kollman, 2004). As Ş tefuriuc (2007: 45) argues "[the territorialisation of electoral competition] requires parties to adapt to a dual logic, as the governing and opposition experiences might overlap in time across levels, and so might the governing-alone and the governing-in-coalition experiences".
In this sense, the Spanish case is particularly interesting for various reasons. On the one hand, regional governments hold significant power and have their own representative institutions. The governing resources and political visibility of the Spanish regions, the so-called Autonomous Communities (Comunidades Autónomas), have progressively expanded as decentralization advanced and, thus, have become crucial arenas for determining and implementing policy (Heller, 2002;Swenden and Maddens, 2009). On the other hand, the coexistence of the centre -periphery cleavage with the ideological left -right axis in some but not all regions has generated significant heterogeneity across the 17 regional party systems (see Wilson, 2012) with significant consequences for government formation. Although regional elections have been classified as second-order elections (Reif and Schmitt, 1980;Jeffrey and Hough, 2003), the dissimilarity of electoral outcomes and the potential (vertical) incongruence of government composition across levels adds intricacy to policy making (Pallarés and Keating, 2003;Hamann and Mershon, 2008).
This article concentrates on SWP, particularly on those which have governing capacity at the central level. SWP compete in all elections and regions throughout the territory and exercise a crucial linkage function between levels of government (Fabre, 2011: 345). Since 1982, only two Spanish SWP have alternated in central government. The PSOE governed from 1982until 1996(1993 and between 2004 and 2011 (in a minority status again), and the PP between 1996 and 2004 (1996-2000 minority government) and from 2011 onwards. Indeed, these two parties account altogether for over 80% of the seats of the lower house. Additionally, the PSOE and the PP are the main parties in most regions and their combined vote share at the regional level has significantly increased over time, nowadays reaching almost 80% (Wilson, 2012: 128). Figure 1 shows the evolution of the regional power hold by the PSOE and the PP. As can be seen, PSOE's predominance was overturned by the PP in 1995. From 1999 until 2011 regional power became well balanced between the main SWPs, with the PSOE taking the lead between 2003 and 2011, when regional power leaned again towards the PP. As of July 2011 the conservatives participated in 11 regional cabinets whereas the PSOE did so in just four-two single-party governments (Andalusia and the Basque Country) and two coalitions in which the Socialists were actually the junior partner and did not hold the premiership (Canary Islands and Navarre).
Throughout the 1980 -2011 period, both SWP were dependent on the support of NSWP at the central level and, at the same time, there were numerous coalitional agreements between SWP and NSWP in the government of several regions. As can be seen in Table 1, from 1980 to July 2011, 30% of all regional governments were coalition governments, either in a minority or a majority status. In 47 out of 55 occasions a SWP was included in the coalition formula (85%). Most often, regional coalition cabinets were composed of a SWP and one or more NSWP (generally one). Interestingly, these governing coalitions very frequently reciprocated support at the central level, where the NSWP provided parliamentary support to the SWP in the Spanish parliament (see Ş tefuriuc, 2009a; Barrio et al., 2010).
Given that SWP -NSWP coalitions have been more frequent for the PSOE than for the PP , we might be inclined to think this is probably due to the fact that the PSOE is closer to the median and/or core position than the PP. However, Figure 2 shows that when the two main dimensions of political competition Figure 1. State-wide parties' presence in regional governments, 1980-2011 (numbers).
are considered (left -right and centralization -decentralization), no clear pattern prevails. A large number of NSWP are either equidistant to both the PSOE and the PP or they are closer to the PSOE on one dimension but closer to the PP on the other. In any case, at the regional level, the PP has also reached numerous coalitional agreements with NSWP, either through participation in a coalition government or through supporting the investiture of a NSWP single-party cabinet.

Theory and Hypotheses
As Ş tefuriuc (2009a: 93-94) argues, decentralized systems come by default with a need to coordinate party action across levels of governance [. . .]. In such settings, political parties operate simultaneously in different party systems, hold different weights therein and need to strike deals with possibly different partners at different levels.
In the same vein, Colomer and Martínez (1995) claim that several parties can agree on simultaneously exchanging their votes in a set of parliaments and, acting rationally in their own self-interest, they can reach coalitional agreements that are apparently paradoxical. This multi-level game is particularly interesting for SWP whose access to a larger set of payoffs means that the losses at one level can be compensated with benefits at the other. As argued by Ş tefuriuc (2009a: 98), for instance, SWP are also those more interested in stepping into congruent coalitions. Vertical congruence refers to the situation in which the party composition of a regional government coincides with that of the central government. SWP will highly value vertical congruence across institutional levels as a means to smooth intergovernmental relations on key policy issues thereby ensuring that regional governments have a fluid relationship with the central executive (Ş tefuriuc, 2009a: 98).
In this article we will focus on the central -regional linkages which work as determinants for SWP's government formation strategies at the regional level using Spain as a case study, one of the most decentralized countries in Western Europe. In particular, we seek to examine the influence of the SWP situation at the central level of government on their choice to strike coalition deals with other allies, basically NSWP, either forming a coalition cabinet or supporting the investiture of a regional prime minister without participating in government.
To study the determinants of SWP's regional government formation strategies, we first need to think about these parties' incentives. It is widely agreed that political parties value office, policy and votes, but different parties might assign a different priority to each of these goals (Strøm, 1990a) and some situations may entail trade-offs between them (Müller and Strøm, 1999). Likewise, in multi-level polities parties might simultaneously pursue different goals at different levels (Downs, 1998). Along these lines, Reniu (2011: 119) argues that SWP and NSWP have different sets of incentives to choose their strategies in the multi-level game of government formation. Access and survival in the central government is an intrinsic goal for SWPs as the polity-wide level is, in Deschouwer's (2003: 217) terms, their "core level of reference". Regional government, then, becomes an instrumental goal for SWPs, but the opposite is true for NSWP-for which the regional level is their core level of reference (see also Ş tefuriuc, 2009b: 100-101). Assuming that (at least the largest) SWP are office-maximizers and that they will prioritize survival in office in the central government, we claim that the behaviour of the party at the regional level will be hierarchically subject to its needs at the central level. In other words, when the position of a SWP at the central level is weak enough to depend on the votes of other parties to gain the investiture, pass policies and/or survive until the next election, then the party's strategies for regional government formation will consider such a situation by engaging in a pattern of vote-exchange across levels.
Hitherto, only single-party governments have formed at the Spanish central level, although majority and minority governments have alternated. In fact, since democratization, the proportion of single-party minority governments has been above 50%. These types of governments have relied on the support of other parties in the Congress of Deputies, very often through stable agreements. On these occasions, stable agreements with parliamentary supporting parties are equivalent to camouflaged or informal governing coalitions (Strøm, 1990b; for an analysis of the Spanish case, see Ş tefuriuc, 2009b). Following the argument above, we expect SWP's strategies at the regional level to be strongly conditioned by the central level when they lead a single-party minority government and need the support of other parties. So we go beyond what has been suggested by previous literature and argue that SWP will not be interested in congruence per se but instead they will engage in a vote-exchange pattern across levels to ensure the survival of the central government when they are in a minority situation there. In particular, we expect a more frequent use of co-operative strategies under these circumstances. Hence, our first hypothesis is the following: Hypothesis 1: In minority situations at the central level, SWP will be more likely to pursue 'coalition' and 'supporting' strategies as opposed to 'singleparty' and 'opposition' ones at the regional level.
Although the above hypothesis may apply to SWP's general strategies at the regional level, the asymmetrical elements of the Spanish political system lead us to think that notable differences will exist across regions (see Ş tefuriuc, 2009a: 99-100). Apart from asymmetry in their levels of fiscal autonomy and differences in their constitutional tracks to autonomy (fast-track versus slow-track regions), regions do also diverge with regard to the strength of their NSWP (Wilson, 2012). Besides the larger SWP (the PSOE, the PP, and the smaller United Left/Izquierda Unida, IU), three NSWP have traditionally been well represented in the lower chamber of the Spanish Parliament (Congreso de los Diputados): the Basque Nationalist Party (PNV, from the Basque Country), Convergence and Union (CiU, from Catalonia), and Canary Coalition (CC, from Canary Islands) and, more recently, Republican Left of Catalonia (ERC, from Catalonia). These parties are sometimes pivotal for the formation of a parliamentary majority at the central level . Other NSWP, although pretty strong in their respective regional parliaments, are small and infrequently represented in the Spanish lower house. 1 Given the regional differences on the potential influence of NSWP at the central level, we qualify the first hypothesis and state our second one: Hypothesis 2: A minority situation at the central level will make SWP more likely to pursue 'coalition' and 'supporting' strategies as opposed to 'single-party' and 'opposition' ones at the regional level, especially in those regions where NSWP are stronger.
Although we acknowledge that coalition formation is, at least, a two-party game, NSWP's preferences will not be considered in our framework. Obviously, we cannot assume that NSWP will always be ready to craft a coalition or a pact for the whole legislative term when approached by a given SWP. However, as long as NSWP are intrinsically interested in access to regional government, we assume that they will accept an agreement with the SWP that offers the most. It is true that the distance between NSWP and the SWP on the main axis of competition will strongly affect the decision taken by the former. However, in the Spanish case, it is very seldom the case that the NSWP is closer to one of the two SWP in both the ideological and territorial dimensions of political competition, as has been shown in Figure 2. Hence, the fact that most NSWP are equidistant to both SWP will make them willing to accept the proposal of the highest bidder. In other words, the agreement will be closed with the party that demands less. 2 Our argument is that the 'bid' of the SWP is more likely to be highest when it finds itself in a situation of minority at the central level and depends on the support of NSWP. As a result, they will behave more cooperatively, either offering the NSWP good conditions to form a coalition government or will demand little to offer their support for the investiture of a NSWP's regional prime minister.
The degree of political decentralization might also affect the types of agreements that parties reach at the regional level. 3 It is reasonable to think that the higher relevance of regional executives in the design and implementation of policies will increase SWP's incentives to control or participate in more regional cabinets, either to co-ordinate central government policies across the country or to confront the other SWP that leads the central government. In addition, political decentralization tends to consolidate the presence of NSWP in the regional electoral arena (De Winter et al., 2006) thereby expanding the opportunities for vote-exchanges between parties. Although regions do not hold a collective veto power within Spain's institutional framework and play a relatively minor role in the policymaking process at the central level (Aja, 2003), horizontal mechanisms among regional governments and intergovernmental bodies have expanded over time. In particular, the larger and the more substantive are shared-competences, the more vertical congruence will matter for SWP (Ş tefuriuc, 2009a: 98). In sum, SWP might increasingly seek regional office as a means to co-ordinate the design and implementation of policies while in central government or as a means to co-ordinate opposition strategies to the central government otherwise. Hence, in our final hypothesis we expect that: Hypothesis 3: SWP's choice for co-operative strategies will become more frequent as political decentralization advances.

Data and Methods
The universe of our empirical analysis is Spanish post-electoral regional governments from 1980 to 2011. However, departing from conventional practice in comparative research of national governments, we consider only those cabinets formed immediately after new elections. Despite some cabinet formations of Spanish regional executives occurring during the term (i.e. inter-election formations), most of them are due to rather unclear changes in the party affiliation or parliamentary group membership of concrete legislators. As a consequence, the distribution of seats and bargaining power between parties in parliament does also change and, therefore, new majorities have to sustain new governments. Apart from the difficulty of identifying these sometimes cumbersome changes in legislators' affiliation/membership, inter-election formations have the disadvantage that they are, by definition, highly dependent on the post-election formation that preceded them. We are interested here in studying SWP's regional formation strategies depending on their situation at the central level when the range of choices is not restricted, by construction, by previous formations. This is why we have included only post-election formations in our analyses, which are in fact the great majority of government formations.
Also, under those situations in which a party obtained the absolute majority of seats in parliament, one might intuitively think that the 'single-party' or 'opposition' were the predominant, if not unique, courses of action. Although it is still possible for majority parties to build oversized coalitions or to receive the support of other SWP in the investiture vote, these situations are very rare in the Spanish regions. The two main SWP have never joined an oversized coalition and the support strategy to a governing party that already had the absolute majority of votes in parliament has been chosen only three times. 4 Given the extremely low frequency of these situations and the irrelevance of the investiture vote in these cases, the empirical analysis concentrates on those cases in which no party obtained a majority of seats after elections, which are substantively far more interesting. In fact, it is in these situations when SWP's strategies (namely, our dependent variable) are really relevant.
Applying these criteria, and excluding the few SWP -SWP agreements, 5 the number of government formation opportunities we analyse is 69. However, since we are interested in evaluating SWP's strategic choices in each regional government formation opportunity, we create two observations for each: one for the PSOE and one for the PP. Both parties decide at the same time what strategy to pursue and even negotiate concurrently with the same potential partners in order to obtain their support for a single-party cabinet, for a coalition, or to engage in a vote-exchange process across institutional levels, and this is precisely what we want to study. Hence, the number of observations expands to 137, 69 for the PSOE and 68 for the PP-the difference resulting from the fact that the PP did not obtain parliamentary representation in the Catalan elections of 1980 and, therefore, did not have any strategic choice available.
Following Hamann and Mershon (2008: 118), we define our units of analysis, political parties, as "recognizable teams". If a group of politicians, even if they formally belong to separate parties, contest elections under the same label we treat them as a single party. This applies to the Catalan nationalist NSWP CiU as well as the electoral coalitions the PP participated in during the 1980s and early 1990s (such as Coalición Popular or Coalición Democrática). Different parties are also considered as one organization if they regularly constitute a single parliamentary group. This is the case of the Party of the Catalan Socialists (Partit dels Socialistes de Catalunya, PSC), which organizationally substitutes for the PSOE in Catalonia, and of the Navarrese People's Union (Unión del Pueblo Navarro, UPN), which has done so for the PP in Navarre for several years (see Verge and Barberà, 2009). For the purposes of the empirical analyses, although formally independent parties, the PSC and the UPN are considered part of the PSOE and the PP organizations, respectively, given that they do not compete against each other and they do not form separate parliamentary groups. 6 It is also important to mention that all throughout the paper we understand SWP's regional coalition strategies as if decided by the central party leadership. Given that multi-level electoral politics introduces centrifugal pressures on SWP (see Hough and Jeffrey, 2006), we can imagine a situation in which a regional branch strikes a coalition deal while the central party opposes it. Yet, the two main Spanish SWP are significantly vertically integrated (Thorlakson, 2009;Fabre, 2011). Whereas political decentralization has clearly shaped parties' electoral and executive strategies, their internal decentralization has remained quite modest (Fabre and Méndez-Lago, 2009;Hopkin, 2009). This is why the central level clearly prevails when it comes to government formation strategies. To the best of our knowledge, in the past 30 years, there have been very few cases in which the central party has de-authorized the coalition agreement reached by a regional party branch, which seems to imply that regional governing strategies are either decided at the central level or at least consensually agreed between party levels. 7

Dependent Variable
Our main dependent variable is SWP's government formation strategy at the regional level. Following regional elections, SWP have various options or strategies at their disposal. As described in Table 2, we have created a nominal variable with four categories, namely single-party, coalition with a NSWP, supporting party (to a NSWP), and opposition, which correspond to the strategies available to SWP. The SWP votes against the investiture of another party's regional prime minister.
The supporting-party role is defined as the SWP voting in favour or abstaining in the investiture of a regional government that does not include itself. Even if voting 'aye' sends a stronger signal of support, abstaining typically allows the investiture of a regional prime minister and her government. This is why we consider abstention in the investiture vote as a 'supporting strategy'. We admit that a truly supporting attitude towards the government from the opposition benches arguably entails more than casting a vote on the single occasion of the investiture. However, the analysis of the complete voting pattern in parliament for three decades would require regional parliamentary voting records that are frequently unavailable. Moreover, our dependent variable refers to government formation strategies rather than SWP's behaviour during the term. Therefore, we have opted for the investiture vote as a highly symbolic gesture reflecting the (un)cooperative intentions of the SWP toward the newly invested government. It bears mentioning that we have also considered as an 'opposition' strategy those abstentions cast in those few regional parliaments where it is not possible to openly vote against the investiture of a candidate as in these case abstention is equivalent to a negative vote.

Independent Variables
As stated in the hypotheses, our main purpose is to identify to what extent multi-level dynamics matter for SWP's strategies in regional government formation. More specifically, we seek to gauge whether or not the course of action Spanish SWP decide to take in different regions at the time to form a (post-election) government is hierarchically subject to the situation of the SWP at the central level (namely, in the Spanish parliament and, subsequently, in the Spanish government). 8 As a result, there are a number of independent variables.
. Central government (CG): This is a categorical variable with three values distinguishing whether (i) the SWP controls the majority of seats at the central level, whether (ii) it governs in a minority situation, or rather (iii) it remains in opposition, the latter being the reference category. . % seats NSWP: This variable measures the percentage of seats obtained by the largest NSWP in each region. It is also interacted with the SWP's status at the central government. . Regional Authority Index: We use the index created by Hooghe et al. (2010), which captures the level of authority of regional entities in decentralized countries on an annual basis. This index measures both the self-rule and shared-rule dimensions of political decentralization. Data are available for the Spanish Autonomous Communities for the period 1978 -2006. As a result, we lose some observations in those analyses where this variable is included, specifically the 2007 and 2011 post-electoral government formations. . Party: This dummy identifies the SWP party in the analyses ('1' for the PP and '0' for the PSOE). Hence, the specified statistical models are party fixed-effect regressions. . Days to general elections: We include as a control variable the time left to the next general elections, measured in days, taking as a benchmark the last regional elections which produced the new (post-election) government. The influence of the situation of the SWP in central government on how it will behave in the regional government might be considerably conditioned by how long the central government is expected to last. Electoral cycles are crucial as the timing of elections at the different levels forces parties to introduce different time horizons simultaneously in their calculations (see Däubler and Debus, 2009;Deschouwer, 2009). As a result, when regional elections are close to the upcoming general elections, SWP might use government formation at the regional level as a "testing ground" for vertical congruence at the polity-wide level (Ş tefuriuc, 2009c: 6).
Given the categorical nature of our dependent variable, we run multinomial logistic regressions where the strategy 'opposition' is specified as the reference category. Since we have two observations for each post-election government formation and the strategies of both the PSOE and the PP are likely to be closely linked, we have specified government-clustered standard errors. 9

Discussion of Findings
According to our first hypothesis, we expect that SWP's situation at the central level will affect their government formation strategies at the regional level. We have argued that their behaviour will tend to be more co-operative when the SWP leads the central government in a minority situation in parliament. Table 3 shows that, in the absence of other control variables, the strategies of 'coalition' and 'support' are less likely when the SWP runs a majority government at the central level, although only the latter is statistically significant at a 99% confidence level. Simultaneously, under this situation, SWP seem to be more prone to form single-party governments at the regional level. These first findings are tentatively consistent with the first hypothesis. Nevertheless, under minority situations at the central government, our first analysis does not reveal any relevant effect: the three displayed strategies of the dependent variable seem to be more likely than the reference category (an overtly opposition strategy), but the estimations are far from statistical significance.
Yet, when the regional dimension is incorporated into the analysis through the variables size of the NSWP (% seats NSWP) and number of days left to the next general  elections the results substantially change. In Table 4 we can see that the interaction between the minority situation of the SWP at the central level and the percentage of seats of the main NSWP in a given region has a positive sign and reaches standard levels of statistical significance for the strategy 'support' (model 2). This finding is congruent with hypothesis 2. Indeed, under minority situations, SWP seem to be more likely to adopt co-operative strategies at the regional level ('support', in particular), most clearly where NSWP are stronger. Although the variable days to general election never reaches the significance level, the coefficient is consistently negative for the 'coalition' and 'support' strategies, showing that the greater the time remaining to the next general election the less co-operative SWP will behave. To put it differently, SWP will be interested in sending a signal of co-operation to NSWP in the region the closer the elections at the central level are in order to secure their support and form congruent coalitions across levels. Figure 3 shows a linear fit simulation of the predicted probability of the 'support' strategy as the size of the main NSWP in the region increases, for two different situations of the SWP at the central government. In those regions where NSWPs are weak, the probability of adopting a 'support' strategy toward them is logically 0, irrespective of the SWP having a minority status at the central government. However, the multi-level effect clearly emerges where NSWPs are stronger. In these regions, SWPs are much more likely to support the investiture of the candidate of a NSWP when they are in a minority status at the central government than in other situations.
Models 3 and 4 incorporate the Regional Authority Index (RAI) in the regressions. As mentioned earlier, the RAI is only available until 2006, which forces us to work with a sub-sample of the longitudinal data used until now. Interestingly, the inclusion of the RAI reveals that, as political decentralization advanced, the 'support' strategy became more likely. This finding is consistent with our third hypothesis: the increasing relevance of regions in the political landscape of Spain seems to have pushed SWP to be more co-operative at the regional level when it comes the time to form a government. Finally, we should also mention that some statistical differences emerge comparing the behaviour of the two main SWP. All else being equal, the PP seems to have been slightly more inclined to adopt co-operative strategies, while it has been less prone (or less able) to govern alone in the regions. The main intention behind the inclusion of this dummy was, however, to control for potential partisan 'fixedeffects'. These differences do certainly deserve further qualitative insight but they fall beyond the scope of our analysis. In sum, we have provided evidence for the influence of the central-level situation of SWP on their regional government formation strategies, controlling for the specific party we look at.

Concluding Remarks
In Western Europe, the regional tier of government has acquired increasing authority and visibility in past decades. Apart from the implications it has had for party organization, government formation in multi-level polities has required parties to co-ordinate their governing action across levels. This need has been especially pressing for SWP, as their governing and opposition experiences overlap in time across levels. Therefore, the analysis of the central -regional linkages in government formation processes requires special attention. This article takes a step forward in this direction. We have focused on SWP's government formation strategies at the regional level in order to disentangle under what conditions SWP will prefer to pact with NSWP.
Our initial expectation was that minority situations at the central government should lead SWP to adopt more co-operative strategies at the regional level in order to secure the support of other parties, namely NSWP, at the central level. This hypothesis has been to a great extent confirmed by the empirical analysis, although only for regions where NSWP have a significant weight. For arithmetical reasons, this is not particularly surprising as we should expect SWP to close agreements with NSWP precisely in those regions where the latter are important political actors. Yet, our results also show that in these regions SWP opt for co-operative strategies (outside parliamentary support, mainly) when they have a minority status at the central government much more frequently than in any other situation. Therefore, our empirical analysis of regional government formation processes in Spain for the entire democratic period (1980 -2011) seems to reveal that multi-level dynamics do matter to understand the strategies that SWP adopt when regional elections have been held and a new government has to be formed. Finally, the empirical analysis has also revealed a dynamic trend. SWP's strategies seem to have changed over time in response to the process of political decentralization. As the regional tier acquired more authority, SWP have become more likely to adopt co-operative strategies.
With a view to confirming the importance of the multi-level dynamics and identifying potential differences across parties as regards their strategies, future qualitative analyses could examine in greater detail the regional government strategies of the PSOE and the PP separately. In-depth investigation of the impact of political decentralization processes on parties' incentives and strategies at the regional level could also prove fruitful. Last, given their simplicity and generalizability, the hypotheses presented in this article can be applied easily to other decentralized countries where the multi-level context makes parties' governing and opposition experiences overlap across tiers of government. 6 As an exception to this general rule, the UPN and the PP are considered as different parties in the Navarrese regional elections of 1983, 1987, and 2011, since both parties then run in separate lists. 7 In 2007 PSOE's regional branch in Navarre was about to close a coalition agreement with several NSWP when negotiations were abruptly called off by the central party leadership (see Fabre, 2008). Regarding IU, although we do not examine this party here, after the 2011 regional elections the Extremadura regional branch contravened the central party instructions to support the investiture of the PSOE candidate and supported instead the investiture of the PP candidate through an abstention. 8 It is important to note that it is beyond the scope of this paper to adopt an integral approach to government formation in Spanish regional governments. There is certainly a greater number of variables at play to understand what types of government finally form. However, as already highlighted, in this article we are just interested in exploring how SWP's situation at the central level affects their government formation strategies at the regional level. 9 We have run additional analyses, including some control variables, to capture the influence of certain institutional factors that may impact on the choice of the government strategies, such as the percentage of deputies needed to present a vote of no-confidence or the consequences of a failed investiture. However, since these variables did not influence the direction or the strength of the coefficients, they are not presented in the tables of results in favour of a more parsimonious specification of the models and a simpler presentation.